Final Project: Crypto Currency from 2013 to 2021¶

Member of our team:

Fullname Student ID
Đoàn Ánh Dương 20127474

Import Library¶

Necessary file imported here.

Note: Some external library out of min_ds-env need to be installed:

  • plotly
In [1]:
import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
from plotly.subplots import make_subplots
import plotly.graph_objects as go

from datetime import datetime, timedelta
from dateutil import relativedelta

import os
import glob

Environment using within this final project.

In [2]:
import sys
sys.executable
Out[2]:
'/home/duong/anaconda3/bin/python'

Set options here.

In [3]:
pd.set_option('display.max_colwidth', None)
# pd.set_option('display.float_format', '{:.2f}'.format)

I. Idea and Expectation¶

Idea¶

For recent years, we have been heard at least once about crypto, particuarly Bitcoin and its application. Although digital currency brought many controversial opinions up to now, we cannot realize its success and popularity.

So I decided to deep in crypto market and fortunately, I found the historical price data of crypto currency on Kaggle. Although the data does not contain all information about a specific coin, the historical price can bring us to take an overview of digital currency evolution.

Throughout this project, I hope that we can have a perspective objectively and a basic knowledge about digital currency.

Expectation¶

In this project, we will find out the crypto currency market and its history, with expectation that we can find a method to buy a coin and take profit.

II. Datasets and Collection¶

Source and Public Data¶

I found the datasets used in this project was public on Kaggle: Crypto Currrency price history.

License¶

The data is taken from coinmarketcap and it is free to use the data.

More about the license I found on kaggle:

  • CC0: Public Domain (No copyright)

About Data¶

The dataset has one csv file for each currency. Price history is available on a daily basis from *April 28, 2013*. This dataset has the historical price information of some of the top crypto currencies by market capitalization, including Bitcoin, Etherium, Binance Coin,...

Description¶

All information of columns contain in the file named description.txt.

Author¶

Datasets were collected by Sudalai Rajkumar(registered account name SRK) on Kaggle.

How did the author collect data?¶

As the author's answer, there is not mentioned clearly how the way he collected data, maybe collected by scraping historical data web. For instance: bitcoin.

Is dataset subjective or objective?¶

Besides the historical price of digital currency, there are many factors in a coin we need to take consideration on. For example: Holders Statistic over time, legality, team and organization investors, who created the coin...

So that the dataset is subjective.

III. Preprocessing Data¶

1. Change files.¶

First, we cannot access each files to analyze, it must be combined into one.

How many files does we have?¶

In [4]:
path = 'archive/'
os.chdir(path)
ext = 'csv'
all_files = [f for f in glob.glob(f'*.{ext}')]

print(f'Containing number of files: {len(all_files)}.')
print(all_files)

os.chdir('../')

Containing number of files: 23.
['coin_XRP.csv', 'coin_Ethereum.csv', 'coin_Aave.csv', 'coin_USDCoin.csv', 'coin_Solana.csv', 'coin_EOS.csv', 'coin_WrappedBitcoin.csv', 'coin_Cardano.csv', 'coin_NEM.csv', 'coin_Tron.csv', 'coin_CryptocomCoin.csv', 'coin_BinanceCoin.csv', 'coin_Iota.csv', 'coin_ChainLink.csv', 'coin_Monero.csv', 'coin_Dogecoin.csv', 'coin_Tether.csv', 'coin_Bitcoin.csv', 'coin_Cosmos.csv', 'coin_Litecoin.csv', 'coin_Stellar.csv', 'coin_Polkadot.csv', 'coin_Uniswap.csv']

Are there duplicated in each files?¶

Lets check duplicated rows.

In [5]:
for f in all_files:
    temp = pd.read_csv(path+f)
    print(f'duplicated in {f}:', temp.index.duplicated().sum())

duplicated in coin_XRP.csv: 0
duplicated in coin_Ethereum.csv: 0
duplicated in coin_Aave.csv: 0
duplicated in coin_USDCoin.csv: 0
duplicated in coin_Solana.csv: 0
duplicated in coin_EOS.csv: 0
duplicated in coin_WrappedBitcoin.csv: 0
duplicated in coin_Cardano.csv: 0
duplicated in coin_NEM.csv: 0
duplicated in coin_Tron.csv: 0
duplicated in coin_CryptocomCoin.csv: 0
duplicated in coin_BinanceCoin.csv: 0
duplicated in coin_Iota.csv: 0
duplicated in coin_ChainLink.csv: 0
duplicated in coin_Monero.csv: 0
duplicated in coin_Dogecoin.csv: 0
duplicated in coin_Tether.csv: 0
duplicated in coin_Bitcoin.csv: 0
duplicated in coin_Cosmos.csv: 0
duplicated in coin_Litecoin.csv: 0
duplicated in coin_Stellar.csv: 0
duplicated in coin_Polkadot.csv: 0
duplicated in coin_Uniswap.csv: 0

So there is no duplicated rows. We carry out combining files.

In [6]:
combined_csv = pd.concat([pd.read_csv(path + fname) for fname in all_files])
print('duplicated in combine files: ', combined_csv.index.duplicated().sum())

duplicated in combine files:  34091

Why the combined files take duplicated? Because column 'SNo' is the index starting with $1$ for each files. So we must reset the index in combined files.

In [7]:
combined_csv.drop(columns='SNo',inplace = True)
combined_csv.to_csv('crypto_finance.csv', index=False)

combined_csv = pd.read_csv('crypto_finance.csv')
combined_csv.index.rename('SNo', inplace = True)
combined_csv.to_csv('crypto_finance.csv')

2. Pre-processing¶

Read new file. Let's see some values.

In [8]:
cryp_df = pd.read_csv('crypto_finance.csv', index_col='SNo')

cryp_df.sample(10)
Out[8]:
Name Symbol Date High Low Open Close Volume Marketcap
SNo
18232 Chainlink LINK 2018-03-21 23:59:59 0.455803 0.390060 0.392459 0.445421 1.282360e+07 1.558974e+08
31198 Litecoin LTC 2013-12-31 23:59:59 24.603800 23.709999 24.505100 24.347000 7.446870e+06 5.952194e+08
22268 Dogecoin DOGE 2014-08-03 23:59:59 0.000210 0.000203 0.000209 0.000203 1.648620e+05 1.803561e+07
31425 Litecoin LTC 2014-08-15 23:59:59 5.212860 4.967520 5.059370 5.048590 1.986680e+06 1.573628e+08
35994 Stellar XLM 2020-03-18 23:59:59 0.037372 0.034913 0.036742 0.035979 3.370208e+08 7.291712e+08
24076 Dogecoin DOGE 2019-07-16 23:59:59 0.003125 0.002802 0.003093 0.002825 4.098083e+07 3.400399e+08
25391 Tether USDT 2016-10-16 23:59:59 1.000000 1.000000 1.000000 1.000000 2.921960e+05 6.951591e+06
24841 Tether USDT 2015-04-15 23:59:59 1.000000 1.000000 1.000000 1.000000 4.070020e+04 2.516000e+05
22169 Dogecoin DOGE 2014-04-26 23:59:59 0.000484 0.000461 0.000476 0.000472 9.989290e+05 3.509400e+07
18123 Chainlink LINK 2017-12-02 23:59:59 0.222175 0.167361 0.177941 0.207898 7.396790e+06 7.276430e+07

More information about dataset.

In [9]:
cryp_df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 37082 entries, 0 to 37081
Data columns (total 9 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   Name       37082 non-null  object 
 1   Symbol     37082 non-null  object 
 2   Date       37082 non-null  object 
 3   High       37082 non-null  float64
 4   Low        37082 non-null  float64
 5   Open       37082 non-null  float64
 6   Close      37082 non-null  float64
 7   Volume     37082 non-null  float64
 8   Marketcap  37082 non-null  float64
dtypes: float64(6), object(3)
memory usage: 2.8+ MB

How many rows and columns in the file?¶

In [10]:
cryp_df.shape
Out[10]:
(37082, 9)

There are 37082 rows and 9 columns.

What is the meaning of each row (sample)?¶

The dataset shows that each samples contain the information of a specific coin at each time.

Does the data have duplicated samples?¶

In [11]:
cryp_df.index.duplicated().sum()
Out[11]:
0

There is no duplicated samples.

What does each column mean?¶

In [12]:
f = open('description.txt', 'r')
print(f.read())
f.close()

1) SNo:         Serial Number.
2) Name:        Name of coin.
3) Symbol:      Symbol of coin.
4) Date:        Date of observation.
5) Open:        Opening price on the given day.
6) High:        Highest price on the given day.
7) Low:         Lowest price on the given day.
8) Close:       Closing price on the given day.
9) Volume:      Volume of transactions on the given day (in USD).
10) Market Cap: Market capitalization in USD.

Check the data types?¶

In [13]:
cryp_df.dtypes
Out[13]:
Name          object
Symbol        object
Date          object
High         float64
Low          float64
Open         float64
Close        float64
Volume       float64
Marketcap    float64
dtype: object

Overall, all the column types is correct except Date. We need to change Date column to appropriate type (datetime).

In [14]:
cryp_df['Date'] = pd.to_datetime(cryp_df['Date'], format="%Y-%m-%d %X")

With the numeric columns, how are values distributed? Is there any abnormal value?¶

In [15]:
nume_cols_df = cryp_df.describe(include=['number', 'datetime'], datetime_is_numeric=True)
nume_cols_df.loc['missing_ratio'] = cryp_df.agg(lambda x: x.isna().mean()*100)

nume_cols_df
Out[15]:
Date High Low Open Close Volume Marketcap
count 37082 37082.000000 37082.000000 37082.000000 37082.000000 3.708200e+04 3.708200e+04
mean 2018-08-16 07:12:30.221616896 1016.058015 952.987707 985.323755 987.120511 3.022542e+09 1.542943e+10
min 2013-04-29 23:59:59 0.000089 0.000079 0.000086 0.000086 0.000000e+00 0.000000e+00
25% 2017-03-05 23:59:59 0.075664 0.069536 0.072456 0.072648 4.937190e+06 2.395955e+08
50% 2019-01-09 23:59:59 1.008733 0.999850 1.001157 1.001138 8.512805e+07 1.405335e+09
75% 2020-05-13 23:59:59 31.916399 28.996246 30.459673 30.512205 9.388489e+08 5.159305e+09
max 2021-07-06 23:59:59 64863.098908 62208.964366 63523.754869 63503.457930 3.509679e+11 1.186364e+12
std NaN 5249.503670 4907.932082 5088.101367 5093.703878 1.190963e+10 7.059128e+10
missing_ratio 0.0 0.000000 0.000000 0.000000 0.000000 0.000000e+00 0.000000e+00

There is no missing value in DataFrame. But it exists abnormal values in min row of Volume and Marketcap col.

So, lets take a look at these columns.

In [16]:
check_df = cryp_df[(cryp_df['Marketcap'] == 0) | (cryp_df['Volume'] == 0)]
check_df
Out[16]:
Name Symbol Date High Low Open Close Volume Marketcap
SNo
0 XRP XRP 2013-08-05 23:59:59 0.005980 0.005613 0.005875 0.005613 0.000000e+00 4.387916e+07
1 XRP XRP 2013-08-06 23:59:59 0.005661 0.004629 0.005637 0.004680 0.000000e+00 3.659101e+07
2 XRP XRP 2013-08-07 23:59:59 0.004682 0.004333 0.004669 0.004417 0.000000e+00 3.453412e+07
3 XRP XRP 2013-08-08 23:59:59 0.004424 0.004175 0.004397 0.004254 0.000000e+00 3.325863e+07
4 XRP XRP 2013-08-09 23:59:59 0.004367 0.004253 0.004257 0.004291 0.000000e+00 3.354750e+07
... ... ... ... ... ... ... ... ... ...
36477 Polkadot DOT 2020-08-28 23:59:59 6.333746 5.540963 5.639486 6.159955 7.271622e+08 0.000000e+00
36478 Polkadot DOT 2020-08-29 23:59:59 6.562906 6.042309 6.175925 6.159143 5.272900e+08 0.000000e+00
36479 Polkadot DOT 2020-08-30 23:59:59 6.219506 5.749978 6.153440 5.869881 4.853351e+08 0.000000e+00
36480 Polkadot DOT 2020-08-31 23:59:59 6.459377 5.772966 5.905918 6.300020 5.126048e+08 0.000000e+00
36481 Polkadot DOT 2020-09-01 23:59:59 6.838898 6.172324 6.298810 6.288767 6.015274e+08 0.000000e+00

971 rows × 9 columns

We got some problems here: Abnormal values on Volume and Marketcap col. We can calculate the ratio of abnormal values to decide whether to drop or not.

In [17]:
abn_df = check_df.groupby(['Symbol']).size()

all_df = cryp_df.groupby(['Symbol']).size()

(abn_df/all_df *100).dropna()
Out[17]:
Symbol
AAVE     0.363636
ATOM     5.443787
BTC      8.090939
CRO      0.213904
DOGE     0.398551
DOT      3.750000
EOS      0.068213
LTC      8.090939
SOL     11.504425
TRX      1.005747
USDC     0.798403
WBTC    22.072072
XRP      4.977532
dtype: float64

As the results, the number of invalid values is quite small, we can delete samples containing invalid values. But once we dropped it, we will face the intermittence of time data. So I decided not to drop.

With the categorical columns, how are the values distributed? Is there any abnormal value?¶

In [18]:
cate_cols = cryp_df.describe(include=[object]).columns

cate_cols_df = cryp_df[cate_cols].agg([
    lambda x: x.isna().mean() *100, 
    pd.Series.nunique, 
    pd.Series.unique])

cate_cols_df.rename(index ={
    '<lambda>': 'missing_ratio',
    'nunique':'num_dif_val',
    'unique':'diff_vals'
})
Out[18]:
Name Symbol
missing_ratio 0.0 0.0
num_dif_val 23 23
diff_vals [XRP, Ethereum, Aave, USD Coin, Solana, EOS, Wrapped Bitcoin, Cardano, NEM, TRON, Crypto.com Coin, Binance Coin, IOTA, Chainlink, Monero, Dogecoin, Tether, Bitcoin, Cosmos, Litecoin, Stellar, Polkadot, Uniswap] [XRP, ETH, AAVE, USDC, SOL, EOS, WBTC, ADA, XEM, TRX, CRO, BNB, MIOTA, LINK, XMR, DOGE, USDT, BTC, ATOM, LTC, XLM, DOT, UNI]

There is no missing value in categorical columns. Everything seems to be fine.

IV. Data Exploratory and Analysis.¶

Data Exploratory¶

which coin is stable/unstable?¶

Before investing any markets, investors must know the basis to save their money. One of the most crucial knowledge is stable/unstable coins:

  • A stablecoin is a digital currency that is pegged to a "stable" reserve asset like the U.S dollar.
  • A unstable coin is a digital currency that has high price volatility, high inflation and weak economy.

We can find the answer by calculating the volatility.

Middle-term Investor (Month)¶
  1. Preprocess

Now we are carrying out calculating the volatility of each coin month by month.

\begin{equation} Monthly\_volatility = \frac{Values(Close\_end\_of\_month)-Values(Close\_month\_before)}{Values(Close\_month\_before)}*100 \end{equation}
In [19]:
last_dom_prep = cryp_df.reset_index('SNo').drop(columns=['SNo','Symbol'])
last_dom_prep['Month'] = last_dom_prep.Date.dt.month
last_dom_prep['Year'] = last_dom_prep.Date.dt.year

last_dom = last_dom_prep.groupby(['Name','Year','Month'])['Date'].max().reset_index()

#make condition
Name = (cryp_df['Name'].isin(last_dom['Name']))
date = (cryp_df['Date'].isin(last_dom['Date']))

#gat all cols of last date each month and each Name.
last_dom_df = last_dom_prep[Name & date]
last_dom_df=last_dom_df.drop(columns=['Month','Year'])

#find min date and max date of each coin => delete min date and duplicate max date.
min_max = last_dom_df[['Name','Date']].groupby('Name').agg([min,max])
min_max = min_max.reset_index(col_level=1).droplevel(level=0,axis=1).set_index(['Name','min'])

substr = last_dom_df.set_index(['Name','Date']).drop(index=min_max.index)

min_max = min_max.reset_index().set_index(['Name','max'])
besubstr = last_dom_df.set_index(['Name','Date']).drop(index=min_max.index).reset_index()

#increase to end of next month
besubstr = besubstr.assign(
    Date = besubstr.Date + pd.Timedelta(days=1)
    )

besubstr = besubstr.assign(
    Date = besubstr.Date.apply(
        lambda x:x+ relativedelta.relativedelta(months=1)
        )
    )

besubstr = besubstr.assign(
    Date = besubstr.Date - pd.Timedelta(days=1)
    )
besubstr = besubstr.set_index(['Name','Date'])

#calculating the volatility
res = ((substr - besubstr)*100/besubstr).reset_index()
res = pd.pivot(res, columns = 'Name',index='Date',values='Close')

res
Out[19]:
Name Aave Binance Coin Bitcoin Cardano Chainlink Cosmos Crypto.com Coin Dogecoin EOS Ethereum ... NEM Polkadot Solana Stellar TRON Tether USD Coin Uniswap Wrapped Bitcoin XRP
Date
2013-05-31 23:59:59 NaN NaN -7.194245 NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2013-06-30 23:59:59 NaN NaN -25.105428 NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2013-07-31 23:59:59 NaN NaN 9.808100 NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2013-08-31 23:59:59 NaN NaN 27.580366 NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2013-09-30 23:59:59 NaN NaN -1.736244 NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN 94.118035
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
2021-04-30 23:59:59 17.962651 106.605778 -1.983501 13.340260 29.904701 18.930354 -8.210962 529.105892 33.862844 44.561196 ... -7.944032 -1.313496 120.680681 28.870459 43.241110 -0.001135 0.021877 45.418800 -2.008534 177.358124
2021-05-31 23:59:59 -13.861344 -43.223781 -35.354564 28.860927 -15.840958 -38.427119 -36.641910 -3.481047 3.160672 -2.100881 ... -44.645530 -35.877113 -23.604979 -23.817326 -41.827906 0.052419 0.044035 -30.291554 -35.398383 -34.246312
2021-06-30 23:59:59 -34.639432 -14.402834 -6.139409 -20.616898 -39.239339 -14.224586 -8.173372 -21.974458 -37.777659 -16.221234 ... -31.126707 -29.997673 8.350011 -29.853028 -11.406481 -0.031692 -0.017570 -32.045981 -6.154983 -32.506669
2021-07-06 23:59:59 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2021-07-31 23:59:59 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN

102 rows × 23 columns

  1. Visualization

Visualizing the volatility of each coin to observe clearly. From now on, we can assume a coin is stable if its volatility is not beyond to $[-15,15]$ and the data is enough for our decisions.

In [20]:
# fig, axs = plt.subplots(int(res.columns.size/2 + 1),2, figsize=(15,60))
# fig.tight_layout()
# for i in range(res.columns.size):
#     axs[int(i/2),i%2].plot(res[res.columns.values[i]])
#     axs[int(i/2),i%2].set_title(res.columns.values[i])

fig = make_subplots(rows = int(res.columns.size/2 + 1), cols = 2,
                    vertical_spacing=0.02,
                    subplot_titles=tuple(res.columns.values))

for i in range(res.columns.size):
    fig.add_trace(
        go.Scatter(x = res.index, 
                    y =res[res.columns.values[i]], 
                    mode='lines',
                    name = res.columns.values[i]  
                ),
        row = int(i/2)+1, col = i%2 + 1
        )

fig.update_layout(height=4000,
                showlegend=False,
                title_text="Volatility of each coin")
fig.show()
Long-term Investor (Year)¶

We could take a look at the vilatility year by year if we are a long-term investor.

All processes like above. We appoximates the yearly volatility base on monthly volatility.

Standard Deviation of monthly volatility.

\begin{equation} STD\_Monthly\_volatility=\sqrt{\frac{\sum{(vola\_month_i- \mu)^2}}{N=Num\_of\_month}} \end{equation}

Anually volatility.

\begin{equation} Annual\_volatility=STD\_Monthly\_volatility \sqrt{Num\_of\_month} \end{equation}
  1. Pre-processing
In [21]:
res1 = ((substr-besubstr)*100/besubstr).reset_index()

#groupby year
res1['Year'] = res1.Date.dt.year
res1 = res1[['Name','Year','Close']].groupby(['Name','Year']).agg([np.std,np.size])\
                                                            .reset_index(col_level=1)
res1 = res1.droplevel(level=0,axis=1)

#Monthly volatility = standard deviation of month-base volatility.
#Annual volatility = Monthly Volatility * sqrt(num time period)
res1['AnnualVola'] = res1['std']*np.sqrt(res1['size'])
res1 = pd.pivot(res1, columns = 'Name',index='Year',values='AnnualVola')
res1
Out[21]:
Name Aave Binance Coin Bitcoin Cardano Chainlink Cosmos Crypto.com Coin Dogecoin EOS Ethereum ... NEM Polkadot Solana Stellar TRON Tether USD Coin Uniswap Wrapped Bitcoin XRP
Year
2013 NaN NaN 457.109181 NaN NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN 539.563746
2014 NaN NaN 66.556166 NaN NaN NaN NaN 362.003329 NaN NaN ... NaN NaN NaN 162.539517 NaN NaN NaN NaN NaN 199.281864
2015 NaN NaN 66.726704 NaN NaN NaN NaN 86.039260 NaN 59.311735 ... 96.102720 NaN NaN 58.689299 NaN 0.000000 NaN NaN NaN 100.429210
2016 NaN NaN 49.309806 NaN NaN NaN NaN 95.502047 NaN 236.411589 ... 600.502428 NaN NaN 93.278866 NaN 0.006333 NaN NaN NaN 65.282208
2017 NaN 1955.111950 95.957177 206.464997 290.086523 NaN NaN 410.123269 320.308490 262.211431 ... 459.309983 NaN NaN 688.910700 1892.737961 13.311229 NaN NaN NaN 869.089489
2018 NaN 73.193491 70.876683 158.615169 147.557427 NaN NaN 151.124506 222.328876 124.953128 ... 122.565094 NaN NaN 158.527085 203.551788 3.668578 0.795658 NaN NaN 132.638788
2019 NaN 119.253488 77.494807 85.825074 267.470088 88.550594 272.493451 56.080850 113.331961 88.975236 ... 103.686407 NaN NaN 69.377906 85.089444 2.223591 2.669780 NaN 76.247803 62.627851
2020 142.704238 72.601705 79.796929 141.519815 156.576276 127.325853 109.703283 76.319909 97.297391 103.716362 ... 192.908441 89.340022 242.260489 187.488532 85.112881 1.125439 3.647628 98.530297 81.547819 204.431611
2021 287.504510 430.664039 74.921018 318.187039 135.334441 150.046174 157.141160 901.397567 78.032515 98.454019 ... 215.813788 164.222431 266.540549 173.644518 141.507427 0.321865 0.118152 285.846007 75.285197 252.159734

9 rows × 23 columns

  1. Visualization
In [22]:
# fig, axs = plt.subplots(int(res1.columns.size/2 + 1),2, figsize=(15,60))
# fig.tight_layout()
# for i in range(res1.columns.size):
#     axs[int(i/2),i%2].plot(res1[res1.columns.values[i]])
#     axs[int(i/2),i%2].set_title(res1.columns.values[i])

fig = make_subplots(rows = int(res1.columns.size/2 + 1), cols = 2,
                   vertical_spacing=0.02,
                    subplot_titles=tuple(res1.columns.values))

for i in range(res1.columns.size):
    fig.add_trace(
        go.Scatter(x = res1.index, 
                    y =res1[res1.columns.values[i]], 
                    mode='lines',
                    name = res1.columns.values[i]  
                ),
        row = int(i/2)+1, col = i%2 + 1
        )

fig.update_layout(height=4000,
                showlegend=False,
                title_text="Volatility of each coin")
fig.show()
Conclusion¶

Before deciding to invest a digital currency, investors must save their capital. How to save their own capital while almost digital currencies have a high volatility? $\Rightarrow$ They must find and reserve in a stable coin.

According to the result, we can consider which one is a stablecoin. Here we have the coins take least volatility whether investors is middle-term or long-term:

  • Tether.
  • USD Coin.

Crypto currency proportion of 2021¶

To see crypto currency evolution, we could use MarketCap col.

In [23]:
true_mcap = cryp_df[(cryp_df['Marketcap'] > 0) & (cryp_df['Date'].dt.year == 2021)]

mean_mcap = true_mcap[['Name','Symbol','Marketcap']].groupby(['Name','Symbol']).mean()
pd.set_option('display.float_format', '{:.2f}'.format)
mean_mcap
Out[23]:
Marketcap
Name Symbol
Aave AAVE 4342005247.88
Binance Coin BNB 45863429127.74
Bitcoin BTC 850308418885.51
Cardano ADA 36073727647.25
Chainlink LINK 11780371063.65
Cosmos ATOM 3439341205.25
Crypto.com Coin CRO 3452219867.56
Dogecoin DOGE 23253147261.34
EOS EOS 4767997966.08
Ethereum ETH 239910890392.16
IOTA MIOTA 3403257862.25
Litecoin LTC 13150059074.47
Monero XMR 4459267395.45
NEM XEM 2850426717.84
Polkadot DOT 25611960118.95
Solana SOL 6176877049.63
Stellar XLM 9270609681.84
TRON TRX 5192913205.79
Tether USDT 44067450019.42
USD Coin USDC 12663720269.43
Uniswap UNI 12035294126.12
Wrapped Bitcoin WBTC 6764035608.35
XRP XRP 33799135106.02

Visualyzing...

In [24]:
mean_mcap = mean_mcap.reset_index()

fig = px.pie(mean_mcap, values= 'Marketcap', names='Name',
            title="Proportion of currency in 2021")
fig.update_traces(textposition = 'inside',textinfo='percent+label')

Result

  • Bitcoin takes the highest market capitalization (60.6%), and the second is Etherium(17.1%).

As a result, it is said that Bitcoin is the main part of crypto currency, and other coins are assumed to be alternative coins (called altcoins).

Bitcoin History By Year¶

History price chart.

In [25]:
btc_df = cryp_df[cryp_df['Symbol'] == 'BTC'].drop(columns='Marketcap')

fig = make_subplots(cols=1, rows=2,
                shared_xaxes=True,
                vertical_spacing=0.05,
                subplot_titles=('Bitcoin history price','Volume'))
fig.add_trace(go.Candlestick(x=btc_df['Date'], 
                    open=btc_df['Open'], 
                    close = btc_df['Close'], 
                    high=btc_df['High'], 
                    low = btc_df['Low'],
                    ),
            row=1,col=1
)

btc_df = btc_df[btc_df['Volume'] > 0]

fig.add_trace(
    go.Scatter(x=btc_df['Date'], y=btc_df['Volume']),
    row=2, col =1
)

fig.update_layout(height=1000,
                xaxis_rangeslider_visible=False,
                showlegend=False,
                title='Bitcoin analysis')

fig.show()

Propose meaningful question & answer.¶

After preprocessing and exploring, we had an overview about the dataset. Lets turn back to our purpose. There are 2 main questions we must answer:

  • How does the price fluctuations of currencies correlate with each other?
  • Which coins do we need to take consideration?

1. How does the price fluctuations of currencies correlate with each other?¶

With this question and the results of data exploratory, we could know how the digital currency work. The answer of this question tells us we should invest alternative coins or bitcoin.

We choose top_7 coins and stable coins except Bitcoin representing alternative coins.

In [26]:
stb_coin=['USDT','USDC']
mean_mcap = mean_mcap.sort_values('Marketcap',ascending=False)
stb_coin = mean_mcap[mean_mcap['Symbol'].isin(stb_coin)]
alt_and_BTC_coin = mean_mcap[:8]

alt_and_BTC_coins = pd.concat([stb_coin,alt_and_BTC_coin],axis=0)
alt_and_BTC_coins = alt_and_BTC_coins.drop_duplicates()

corr_df = cryp_df[cryp_df['Symbol'].isin(alt_and_BTC_coins['Symbol'])]
corr_df = pd.pivot(corr_df,values='Close', index='Date',columns='Symbol')

corr_df
Out[26]:
Symbol ADA BNB BTC DOGE DOT ETH USDC USDT XRP
Date
2013-04-29 23:59:59 NaN NaN 144.54 NaN NaN NaN NaN NaN NaN
2013-04-30 23:59:59 NaN NaN 139.00 NaN NaN NaN NaN NaN NaN
2013-05-01 23:59:59 NaN NaN 116.99 NaN NaN NaN NaN NaN NaN
2013-05-02 23:59:59 NaN NaN 105.21 NaN NaN NaN NaN NaN NaN
2013-05-03 23:59:59 NaN NaN 97.75 NaN NaN NaN NaN NaN NaN
... ... ... ... ... ... ... ... ... ...
2021-07-02 23:59:59 1.39 287.42 33897.05 0.25 15.34 2150.04 1.00 1.00 0.66
2021-07-03 23:59:59 1.41 298.24 34668.55 0.25 15.55 2226.11 1.00 1.00 0.67
2021-07-04 23:59:59 1.46 307.73 35287.78 0.25 16.01 2321.72 1.00 1.00 0.69
2021-07-05 23:59:59 1.40 302.38 33746.00 0.23 15.24 2198.58 1.00 1.00 0.65
2021-07-06 23:59:59 1.42 320.93 34235.19 0.23 16.14 2324.68 1.00 1.00 0.67

2991 rows × 9 columns

Calculating correlation efficient.

In [27]:
corr_df.corr()
Out[27]:
Symbol ADA BNB BTC DOGE DOT ETH USDC USDT XRP
Symbol
ADA 1.00 0.89 0.86 0.83 0.85 0.96 -0.24 -0.07 0.65
BNB 0.89 1.00 0.86 0.89 0.85 0.90 -0.22 -0.08 0.48
BTC 0.86 0.86 1.00 0.65 0.95 0.90 -0.32 0.04 0.62
DOGE 0.83 0.89 0.65 1.00 0.57 0.84 -0.17 0.00 0.49
DOT 0.85 0.85 0.95 0.57 1.00 0.84 0.02 -0.08 0.71
ETH 0.96 0.90 0.90 0.84 0.84 1.00 -0.27 0.03 0.70
USDC -0.24 -0.22 -0.32 -0.17 0.02 -0.27 1.00 0.23 -0.11
USDT -0.07 -0.08 0.04 0.00 -0.08 0.03 0.23 1.00 0.07
XRP 0.65 0.48 0.62 0.49 0.71 0.70 -0.11 0.07 1.00
Result & Conclusion¶

Result

  • Among Bitcoin and stable coins, we has negative correlation (weak correlation with USDT). And with others coins, it is shown that almost cases have negative correlations (weak correlation with USDT).
  • Among Bitcoin and alternative coins, we has strong positive correlations.

Conclusion

  • As the results showed us, We could choose alternative coins to invest by following Bitcoin. This conclusion implices that we could invest alternative coins to optimize profits and following Bitcoin if its profit is greater.
  • Secondly, we can understand a little bit about other investors' sentiment: if the scenario is bullish market, some investors exchanged from stable coins (USDC) to coins having higher volatility. $\Rightarrow$ stable coins is down and other is up (negative correlation).
Analyze some results¶
In [28]:
eth_df = cryp_df[cryp_df['Symbol']=='ETH']
tether_df = cryp_df[cryp_df['Symbol']=='USDT']
In [29]:
fig = make_subplots(cols=1, rows=3,
                shared_xaxes=True,
                vertical_spacing=0.05,
                subplot_titles=('Bitcoin historical price',
                                'Ethereum historical price',
                                'Tether historical price'
                                ))
fig.add_trace(
    go.Candlestick(x=btc_df['Date'], 
                    open=btc_df['Open'], 
                    close = btc_df['Close'], 
                    high=btc_df['High'], 
                    low = btc_df['Low'],
                    ),
    row=1,col=1
)

fig.add_trace(
    go.Candlestick(x=eth_df['Date'], 
                    open=eth_df['Open'], 
                    close = eth_df ['Close'], 
                    high=eth_df['High'], 
                    low = eth_df['Low'],
                    ),
    row=2, col=1
)

fig.add_trace(
    go.Candlestick(x=tether_df['Date'], 
                    open=tether_df['Open'], 
                    close = tether_df['Close'], 
                    high=tether_df['High'], 
                    low = tether_df['Low'],
                    ),
    row=3, col=1
)

for i in range(3):
    fig.update_xaxes(row=i+1, col=1, rangeslider_visible=False)

fig.update_layout(height=1500,
                showlegend=False,
                title='Crypto analysis')

fig.show()

But, how to choose other coins for optimizing profits. We continue to the next question.

2. Which coins do we need to take considerations?¶

We can answer this question by being greedy in profits... which coins has the best profits.

Do High,Low,Open,Close cols have values been less than or equal to 0?

In [30]:
(cryp_df[['High','Low','Open','Close']] <= 0).sum()
Out[30]:
High     0
Low      0
Open     0
Close    0
dtype: int64
About profits¶

Get max profits by dividing max of high to min of low for each coins.

\begin{equation}Profits\_of\_coins = \frac{Max(High)}{Min(Low)}\end{equation}
In [31]:
calc_coins = cryp_df[['Name','Symbol','High','Low']].groupby(['Name','Symbol']).agg([np.min,np.max])

#Find max of High and min of Low
calc_coins = calc_coins.assign(
    High_max = calc_coins['High']['amax'],
    Low_min = calc_coins['High']['amin']
)
calc_coins = calc_coins.drop(columns=['High','Low']).droplevel(level=1,axis=1)

#Calculating profits
profits = pd.DataFrame(calc_coins['High_max'] / calc_coins['Low_min'])
profits = profits.rename({0:'Proportion'},axis=1)
profits = profits.reset_index().drop(columns='Symbol')

#
profits = profits.sort_values('Proportion',ascending=False)
profits
Out[31]:
Name Proportion
13 NEM 23506.73
9 Ethereum 9032.01
7 Dogecoin 8269.84
1 Binance Coin 6826.65
12 Monero 2056.03
22 XRP 1246.49
2 Bitcoin 869.93
16 Stellar 621.70
4 Chainlink 335.85
11 Litecoin 307.08
17 TRON 164.07
3 Cardano 116.95
15 Solana 104.16
10 IOTA 45.62
8 EOS 44.13
0 Aave 23.45
6 Crypto.com Coin 21.74
20 Uniswap 20.82
21 Wrapped Bitcoin 18.70
5 Cosmos 17.72
14 Polkadot 16.08
18 Tether 2.00
19 USD Coin 1.11

Visualyzing...

In [32]:
plt.figure(figsize=(8,6))
plt.style.use('seaborn-whitegrid')
sns.barplot(x = profits['Proportion'], y = profits['Name'])
plt.title('Profits')
plt.xlabel('Ratio');
Conclusion

We can take consideration on NEM, Ethereum, Dogecoin, Binance Coin, Monero,... before actually investing.

About risks¶

Calculating risk based on the last day of Dataset collected:

\begin{equation} Risks=\frac{min(Low\_of\_last\_day)-max(High)}{max(High)}*100 \end{equation}
In [33]:
max_date = cryp_df[['Name','Symbol','Date']].groupby(['Name','Symbol']).max()\
                            .reset_index().set_index(['Name','Symbol','Date'])

#get last days
max_date_df = cryp_df.set_index(['Name','Symbol','Date'])
max_date_df = max_date_df[max_date_df.index.isin(max_date.index)]\
                        .reset_index().set_index(['Name','Symbol']).drop(columns='Date')

#Take risks
risk = pd.DataFrame(
        (max_date_df['Low'] - calc_coins['High_max'])*100/calc_coins['High_max']
    ).rename({0:'Risks'},axis=1).reset_index()

#sorting values
risk = risk.sort_values('Risks')
risk
Out[33]:
Name Symbol Risks
13 NEM XEM -93.84
10 IOTA MIOTA -85.58
8 EOS EOS -83.41
22 XRP XRP -83.01
17 TRON TRX -78.69
16 Stellar XLM -72.97
14 Polkadot DOT -69.39
7 Dogecoin DOGE -68.84
11 Litecoin LTC -67.09
4 Chainlink LINK -65.29
5 Cosmos ATOM -60.10
12 Monero XMR -58.99
6 Crypto.com Coin CRO -57.29
1 Binance Coin BNB -56.26
20 Uniswap UNI -55.36
0 Aave AAVE -53.81
9 Ethereum ETH -49.62
2 Bitcoin BTC -48.20
21 Wrapped Bitcoin WBTC -47.99
15 Solana SOL -43.52
3 Cardano ADA -43.40
18 Tether USDT -17.53
19 USD Coin USDC -10.24

Visualyzing...

In [34]:
plt.figure(figsize=(8,6))
plt.style.use('seaborn-whitegrid')
sns.barplot(x = risk['Risks'], y = risk['Name'])
plt.title('Risks')
plt.yticks(rotation=45)
plt.xlabel('Percentage');
Conclusion

We can take consideration on NEM, IOTA,EOS, XRP... before actually investing.

Question Conclusion¶

We can easily observe that some coins having very good profits(like NEM can bring to investors $> 20000$ times based on their money in), but besides it, it could take the investors' money by substact $93.84%$ based on their money in.

So investors can pour out their money on the coins that take less risks, E.X: Solana or Bitcoin,...

**Note**:The answer to this question is for reference only

Project Reference¶

  • Documents of subject Code for Data Science - HCMUS.
  • Searching functions on Plotly - Plotly Documentation.
  • Searching functions on Pandas - Pandas Documentation.
  • Searching some financial formula - Wall street.